Examples

1. 1.

   For example, suppose the actual range of values of a certain feature is 800 to 2,400. As part of feature engineering, you could normalize the actual values down to a standard range, such as -1 to +1.

2. 2.

   Normalization is a common task in feature engineering. Models usually train faster (and produce better predictions) when every numerical feature in the feature vector has roughly the same range.

3. 3.

   See also Z-score normalization.

Real-world uses

Created for this library

1. 1.

   A retail data team applies feature normalization at training time and at serving time using stored statistics for consistency.

2. 2.

   An NLP team applies text normalization steps like lowercasing and unicode normalization before tokenizing inputs.

3. 3.

   A computer vision team applies per-channel image normalization at the input layer of its CNN before any convolution.